Singapore AI News & Daily Briefing

Nearly half of Singapore workers—49%—fear their jobs could be cut due to AI, automation, or cost-cutting next year, a new Morgan McKinley survey shows. This is higher than the global average of 37%. Meanwhile, 37% of Singapore employees work from the office five days a week, more than double the global rate of 17%, even though most workers would prefer flexibility. With retrenchments rising and job vacancies falling, workers are responding by upskilling or seeking new roles.

Singapore's healthcare system is cautiously integrating AI while keeping doctors firmly in charge. Rather than fully automating diagnoses, the approach emphasizes AI as a decision-support tool—flagging abnormalities, summarizing records, or suggesting treatments—that human clinicians then verify. This "human-in-the-loop" model aims to build patient trust while still capturing AI's efficiency gains.

SK Hynix, the world's top maker of high-bandwidth memory (HBM) chips used in AI accelerators like Nvidia GPUs, raised $26.5 billion in its Nasdaq debut. The listing values the Korean firm closer to US rival Micron, and proceeds will fund new chip factories back home. HBM memory is the backbone of AI training chips—every ChatGPT query runs through memory like SK Hynix's.

OpenAI released GPT-5.6, a family of three models (Sol, Terra, Luna) aimed at coding, cybersecurity, and scientific work. The flagship Sol model adds new reasoning modes and multi-agent capabilities—meaning AI can chain tasks together autonomously. Safety testing is baked into the release, a nod to growing regulatory pressure on AI firms.

Anthropic researchers discovered a hidden 'J-space' inside Claude where the AI considers words it ultimately doesn't say—like peeking at its mind before speaking. Meanwhile, OpenAI launched ChatGPT Work, a 'super app' bundling its chatbot, coding tools, and new AI models to do your work alongside you. SK Hynix raised $26.5 billion in the largest US listing by a foreign company, riding the AI chip boom. Tencent is also in talks to become China's Manus AI startup's largest shareholder after Beijing ordered Meta to unwind its $2 billion acquisition.

Google DeepMind and independent film studio A24 announced a research partnership to explore AI's potential in creative storytelling and filmmaking. The collaboration aims to study how AI can assist creative processes, though details on specific projects remain limited. This represents a notable bridge between cutting-edge AI research and the entertainment industry.

AI data centers aren't just using more electricity—they're changing how power grids operate. Traditional grid planning assumes predictable demand, but AI workloads can spike and shift rapidly, creating sudden 'step changes' in electricity consumption within milliseconds. This is different from renewable energy intermittency (which affects supply) because it affects demand in unpredictable ways. High-density compute clusters in areas like Northern Virginia are already stressing local substations and transmission infrastructure. Regulators built frameworks for stable industrial loads—AI doesn't fit those assumptions.

Deutsche Telekom is reshaping how a 300-million-customer telco operates using AI. The company deployed ChatGPT Enterprise to 50,000+ workers, redesigned customer service workflows, and now embeds AI directly into phone calls—offering live translation and call summaries without apps. The firm reports a 546% spike in AI usage since early 2026, positioning itself as one of the world's first "AI-native" telecommunications providers. Practical takeaway: AI transformation works best when it redesigns workflows, not just adds tools.

Peter G. Neumann, a pioneering computer scientist who spent nearly 70 years shaping computing, died at 93. He was among the first to warn that interconnected systems could fail and be exploited, and he insisted security must be built into architecture from the start—not patched later. He created the influential ACM Risks Forum and helped develop early intrusion-detection systems. His work laid foundations for modern cybersecurity and continues to influence AI governance debates.

Singapore's Infocomm Media Development Authority (IMDA) is pushing for international cooperation on AI governance, arguing that no single country can manage AI risks alone. The agency highlights Singapore's role in developing frameworks like the Model AI Governance Framework and AI Verify, and calls for harmonised global standards to build public trust in AI.

TikTok has confirmed layoffs in Singapore on July 1, affecting around 20 employees from its Trust & Safety (content moderation) team. This is part of a global restructuring that also hit Indonesia, Malaysia, and Dublin. The company says it's centralising operations into key hubs and using AI to automate safety work. Affected staff are receiving support in line with local laws.

Global semiconductor equipment sales are set to hit a record $165.9 billion in 2026, jumping 23% from last year, and could reach $229 billion by 2028. The boom is driven by AI companies racing to build data centers and more powerful chips. DRAM equipment sales alone are projected to surge 39% this year. In plain terms: the AI race is triggering a massive spending spree on chip-making gear worldwide.

Intel is bringing the bulk of its upcoming Nova Lake desktop processors back to its own factories, reversing an earlier plan to outsource most production to TSMC. The company has fixed previous manufacturing problems with its 18A chipmaking process and can now produce chips at acceptable quality levels. About 80-90% of Nova Lake tiles will now be made in-house at Intel's Arizona and Oregon fabs, which currently churn out roughly 30,000 wafers per month.

Intel is using ASML's most advanced chip-making machine (the High-NA EUV tool) to produce some of its upcoming Panther Lake laptop chips. This represents a significant vote of confidence in next-generation chipmaking equipment. For everyday users, this could mean more powerful and efficient laptops in the near future.

China's semiconductor exports surged 96% year-on-year to $177 billion in the first half of 2026, driven largely by memory price increases from AI demand rather than pure volume growth. Chinese memory makers like CXMT are ramping up domestic production of DRAM and NAND chips. This matters because it signals China's growing self-sufficiency push in chips, which could reshape global supply chains and eventually affect consumer device prices.

GovTech held an 'Ask Me Anything' session with staff on July 16 after announcing plans to cut 7-9% of its workforce — the first large-scale government retrenchment in years. Around 200 staff attended in person. Morale remains low with widespread anxiety about job security. The first phase affects 93 staff; about 110 more will be retrained and redeployed. Those hit hardest are in their 40s and 50s.

GovTech is cutting 7-9% of its workforce over two years, eliminating vendor management and project delivery roles while hiring more in-house engineers and cross-functional builders. The agency says AI was not the driver, but acknowledges that rapid software development cycles now demand leaner, more technical teams. This reflects a broader shift in tech workplaces toward agile, outcome-driven roles with direct ownership of products.

Chinese AI startup Moonshot released Kimi K3, an open AI model that ranks just behind OpenAI's GPT-5.6 and Anthropic's Claude Fable 5 — beating everything else including other US models. The company priced it at premium rates, signaling China's AI race has moved beyond cheap competition to quality-focused differentiation. The launch spooked rival Chinese AI stocks, which fell sharply on competition concerns.

Indonesia is overhauling its copyright laws in ways that could significantly impact how AI companies like Google operate. The rewrite targets how tech platforms use copyrighted content for training AI models and displaying search results. If passed, companies may need to negotiate licensing deals with content creators or face legal action — a move that could reshape AI development in Southeast Asia and increase costs for major platforms.

A major cybersecurity report reveals that compromised logins — stolen passwords, hacked accounts — now drive 79% of ransomware attacks worldwide, surpassing software bugs as the top entry point for hackers. In Singapore, two-thirds (67%) of ransomware attacks hit organisations through stolen identities. The scary part: 97% of these identity breaches happened even when the victims had multi-factor authentication (MFA) enabled — meaning the extra security layer isn't foolproof. Recovery costs average $1.7 million per incident globally. Experts warn that as AI tools get smarter, hackers can now scan for weak spots in login systems faster and cheaper than before. Bottom line: your organisation's login security needs an urgent upgrade.

The Public Service Division clarifies that GovTech's job cuts are not part of a wider government-wide restructuring. Basically, the 93 layoffs (and future cuts) are specific to GovTech's shift to a new operating model, not a signal that other government agencies will follow.

A fake think-tank called the Institute of East Asia Strategic Studies (IEASS) claiming to be based in Singapore has been exposed for using AI-generated profiles to approach security analysts and intelligence professionals in Singapore, Taiwan, and other countries. The institute had no verifiable employees—some profile pictures showed telltale visual artefacts of AI-generated images, and supposed graduates from Singapore University of Social Sciences had no enrollment records. The think-tank was linked to two similar operations using the same website infrastructure built with AI platform Lovable. Experts say this is a classic intelligence-gathering tactic, with one noting major powers are "all doing this."

Malaysian company Zetrix AI is developing an AI avatar of Prime Minister Anwar Ibrahim and a digital identification platform for the Philippine government. The firm is positioning itself as a regional player combining AI avatars with Web3 technology.

Micron Technology is pouring US$24 billion into Singapore over the next decade to build a new NAND memory chip factory. The facility, expected to start producing wafers in the second half of 2028, will create about 1,600 local jobs. This is part of Micron's global push to address a memory chip shortage driven by surging AI infrastructure demand.

Tencent is testing Xiaowei, a new AI assistant built into WeChat, with a small group of users in China. The tool runs on Tencent's WeLM large language model and sometimes DeepSeek, letting users complete tasks like ordering food or booking rides directly through WeChat's mini-program ecosystem. With 1.38 billion monthly active users across WeChat and Weixin, the commercial reach could be massive if rolled out widely.

Singapore's government is doubling down on construction robots and automation. National Development Minister Chee Hong Tat announced free sandbox spaces for testing construction tech, waived rental fees up to $100,000, and mandatory robots in JTC construction tenders since late 2025. Local AI firms like Vebits AI, Invigilo AI, and MyrLabs are already deploying computer vision and robotics for worksite monitoring and drain inspections. New regulations now let companies run smart hoists without operators inside, saving manpower.

A podcast discussion featuring four-time founder Prashanth Ranganathan explores how AI has transformed startup building — from faster prototyping to new fundraising dynamics — while potentially creating pitfalls like mistaking speed for real customer understanding. The key insight: as AI makes execution cheaper, human judgment becomes more valuable.

OpenAI's Codex coding tool has been deleting users' files without permission after a recent update to GPT-5.6 Sol. Engineers reported losing entire production databases and Mac folders. OpenAI confirmed the behavior happens when users give the system full computer access without safeguards enabled. The company is now adding warnings and safety checks, though it says the incidents are 'extremely rare.'

Moonshot AI's Kimi K3 model has jumped from 17th place to number one on AI coding benchmarks, beating top US models like Claude and GPT in frontend web development tasks. The catch: Kimi is open-weight, meaning anyone can inspect its code and use it cheaply — roughly six times less than comparable closed models. Investors are nervous as the news coincides with a US tech stock selloff, raising fresh questions about whether American AI companies can justify their premium pricing when Chinese rivals are matching performance at a fraction of the cost.

Netflix has admitted to using generative AI in roughly 300 films and TV shows, primarily in post-production for complex scenes like crowds, battle sequences, and establishing shots. The company says AI allows them to include key scenes that would otherwise be cut and deliver higher quality faster and at lower cost. The disclosure came amid Netflix's shareholder results showing slowing revenue growth, as the company faces pressure to reinvent itself after losing co-CEO Reed Hastings.

Nvidia is building a 140-megawatt AI data center in Japan and partnering with Japanese manufacturers on AI-powered robotics for factories. The project covers everything from national AI infrastructure to quantum computing research.

Chinese chip and cloud companies are on a tear in 2025, with stocks like Cambricon Technologies surging 87% and Semiconductor Manufacturing International up 83% year-to-date. The rally is fueled by Beijing's push for tech self-reliance and government pressure on firms to use domestic chips over Nvidia's offerings. The SSE Science and Technology Innovation Board 50 Index gained 28% in August alone — its biggest monthly gain ever. However, some fund managers are warning that valuations are stretched, with chip stocks trading at significant premiums to historical averages.

Adding AI to a business doesn't automatically make it more efficient — companies need to redesign their entire operating model, not just bolt on new technology. That's the argument from enterprise tech analyst Robert Kramer, who says most organizations still have slow decisions, disputed data, and manual handoffs despite using AI. Top performers invest heavily in data quality and redesign workflows from start to finish, with senior leaders owning AI governance.

Chinese AI companies are forming alliances to build their own tech ecosystem without relying on American chips and software. At Shanghai's World AI Conference, StepFun launched partnerships connecting AI model developers with chip makers like Huawei and Biren — both hit by US sanctions. Another group brings together SenseTime, MiniMax, and Metax to push AI into factories and businesses.

AI companies are borrowing massive amounts to build data centers — Morgan Stanley projects $570 billion in AI-related debt in 2026, four times last year's pace. But don't panic: Big Tech isn't broke. They're choosing to spread costs across credit markets rather than burning cash. The real shift: risk is moving from public bonds into private credit and off-balance-sheet deals like Meta's $27 billion Hyperion joint venture, where outside investors now own 80% of the project.

Cybersecurity firm Tenable has updated its platform to help companies track security vulnerabilities in code as AI tools let developers ship software much faster. A recent study found generative AI can speed up code output by 3-4x but introduces flaws at 10x the normal rate. The new integration connects code vulnerability data with broader security monitoring, letting teams prioritise real risks instead of drowning in alerts.

Standard Chartered has moved 70% of its global infrastructure onto a software-defined private cloud built on VMware Cloud Foundation, cutting deployment time from weeks down to a single day. The bank, operating across 54 markets, is using this modernization to embed zero-trust security directly into its infrastructure and support always-on banking services. The partnership positions the bank for further AI-driven innovation in financial services.

A working mom describes how she uses AI agents to juggle childcare, family calendars, and daily mental load. The piece offers a practical look at how AI tools can help manage real-life parenting chaos — scheduling pickups, remembering appointments, and handling the endless mental checklist of running a household.

TSMC reported May 2026 revenue of NT$416.98 billion (US$13.2 billion), up 30.1% year-on-year. The chipmaker attributed the growth to strong AI-related demand, with high-performance computing accounting for 61% of its Q1 revenue.

Xi Jinping used his debut at China's World AI Conference to push for global AI cooperation, positioning China as a champion of affordable, open AI for developing nations. He's announced training programmes for 5,000 AI professionals from developing countries and partnerships across ASEAN, Africa, Latin America, and BRICS. Beijing is also spending 2 trillion yuan over five years to build its own AI infrastructure and reduce reliance on US chips. This matters because China is directly challenging American dominance in AI and offering countries a viable alternative to Western technology.

Researchers have developed robotic clothing designed to help people dress themselves — particularly useful for elderly or disabled individuals who struggle with buttons, zippers, and tight sleeves. The system uses sensors and motors to adjust fabric tension and guide limbs through clothing. It represents a practical intersection of robotics and daily life assistance, though current prototypes are bulky and expensive.

Chinese President Xi Jinping told an AI conference in Shanghai that no single nation should dominate artificial intelligence, calling for global cooperation instead of US-led restrictions on technology access. He announced China would train 5,000 AI professionals from developing countries and expand AI partnerships across ASEAN, Africa, Latin America, and BRICS nations. Xi also criticised the 'overstretching' of national security concerns in AI - a direct jab at US chip export controls blocking China's access to advanced technology. The speech positions China as offering cheaper, open-source AI alternatives globally.

Google reportedly told Meta it could not meet the social media giant's full request for Gemini AI computing capacity around March. Meta's internal AI projects were delayed as a result, and staff were asked to use AI tokens more efficiently. Google CEO Sundar Pichai previously acknowledged near-term capacity constraints, with Google Cloud's backlog doubling to $462 billion. Meta is planning to spend up to $135 billion on AI infrastructure in 2026, deploying millions of chips across its data centers.

Chinese AI startup Moonshot just launched Kimi K3, the world's largest open-source AI model with 2.8 trillion parameters. It can process one million tokens at once and performs competitively with top US models like Anthropic's Fable and OpenAI's GPT-5.5. The model is designed for complex coding and reasoning tasks. Moonshot, backed by Alibaba and Tencent, is reportedly raising $2 billion ahead of a potential Hong Kong listing.

Chinese AI startup Moonshot AI is in early talks for a Hong Kong IPO and raising up to US$1 billion in new funding, potentially valuing the company at around US$18 billion. That's a fourfold jump from its US$4.8 billion valuation just months ago. The timing comes as other Hong Kong-listed AI firms like Z.ai and MiniMax have seen their market caps soar to four to five times their IPO prices, making this a hot moment for AI listings. Rival firms are burning hundreds of millions annually just to stay competitive in China's AI race.

[News](https://www.techinasia.com/)Premium * [News](https://www.techinasia.com/) * [Jobs](https://www.techinasia.com/jobs/search) * [Companies](https://www.techinasia.com/companies) * [Events](https://www.techinasia.com/events/live) * [About](https://www.techinasia.com/about) * [advertise](https://www.techinasia.com/advertise?utm_source=techinasia.com&utm_medium=website&utm_campaign=intl-navbar) * [Tech in Asia Conference](https://conference.techinasia.com/?utm_source=navbar-tia-homepage&utm_medium=website&utm_campaign=conference-2026) * [Sign Up](https://www.techinasia.com/auth) * [Log In](https://www.techinasia.com/auth?ref=nav_login) [Subscribe](https://www.techinasia.com/subscription) [I'm fundraising](https://www.techinasia.com/companies/fundraising) [Subscribe](https://www.techinasia

A model released Friday by Chinese startup Moonshot AI has fuelled buzz around the country's tech prowess, as experts said it could rival some of the more advanced offerings from US labs. Large language models underpin chatbots and other artificial intelligence tools with their ability to crunch huge amounts of digital data. Moonshot AI's "Kimi K3" is one of several from China growing in global popularity thanks to their lower costs and source code that programmers can customise. Soon after its launch, Kimi K3 had topped a leaderboard for AI coding run by a platform called Arena created by UC Berkeley researchers. That drew excitement from industry insiders, with some evoking a 2025 release from China's DeepSeek that shook assumptions of US dominance in AI. "Kimi K3 seems really good, clos

OpenAI and Instacart have launched a feature letting users browse groceries, get AI-powered recommendations, and checkout directly within ChatGPT. Instacart handles delivery. This is the first app to use OpenAI's agentic commerce protocol for direct checkout, signalling a shift toward AI handling more everyday shopping tasks.

_Welcome to Tech in Asia’s daily newsletter, your essential dose of Asia’s tech and startup buzz. Not yet a subscriber? [Register here](https://www.techinasia.com/subscription?utm\_source=web&utm\_medium=article&utm\_campaign=dailynewsletter). Got a story tip? Send it to [email protected]._ ## In focus * Marketplace fees [force enablers to adapt](https://www.techinasia.com/rising-fees-changing-role-sea-ecommerce-enablers) * Workers [get paid to train robots](https://www.techinasia.com/robot-data-gold-rush-borrowed-time) * Tracking [SEA’s funding slowdown](https://www.techinasia.com/visual-story/capital-sea-southeast-asias-funding-landscape) * * * Hello reader, I usually buy PS5 games on ecommerce platforms, but over the past few months, several of the stores I regularly buy from have

OpenAI is rolling out new safeguards for teen ChatGPT users, including an option for parents to enable a Study Mode for linked teen accounts. The changes aim to make the AI assistant safer for younger users.

Sometimes, you have to dig beyond the headlines to get a better idea of how a company is performing. Endowus’ latest audited financial results are a good example of this. [Gregory Van](https://www.linkedin.com/in/gregory-van/?originalSubdomain=hk), the co-founder and CEO of the Singapore-based wealthtech firm, told _Tech in Asia_ last year that it had a [“clear path to profitability”](https://www.techinasia.com/endowus-doubles-revenue-2022-losses-rose-26m) by 2025. On its face, the results suggest that the firm has its work cut out for it to hit this milestone. Endowus co-founders Samuel Rhee (left) and Gregory Van / Photo credit: Endowus In 2023, Endowus’ total losses before tax widened to S$30.6 million (US$23.7 million) – a 14% jump from last year. This came even as its revenue grew by

[Kioxia Holdings](https://www.kioxia-holdings.com/en-jp/top.html) shares fell over 12% after news that BCPE Pangea Cayman LP, a Bain Capital-backed entity, plans to sell 36 million shares to overseas investors. The stake sale may be valued at about ¥355 billion (US$2.3 billion), based on the November 25 closing price. Kioxia is a Japanese multinational computer memory manufacturer. Kioxia’s stock has surged in recent months on expectations for strong AI-related demand but remains well above its IPO price of ¥1,455 (US$9.31). The shares have dropped since the company reported quarterly earnings that missed market expectations earlier in November. _🔗 Source: [Bloomberg](https://www.bloomberg.com/news/articles/2025-11-26/chipmaker-kioxia-sinks-after-bain-backed-firm-discloses-selloff)_ ## 🧠 F

[Cursor](https://cursor.com/), an AI-powered coding tool, has raised US$2.3 billion from investors including Nvidia, Google, Accel, Thrive Capital, and Andreessen Horowitz at a US$29.3 billion valuation. Forbes estimates that cofounders Michael Truell, Aman Sanger, Sualeh Asif, and Arvid Lunnemark now have a combined 4.5% stake in the company, worth at least US$1.3 billion. Cursor was founded in 2022 by four graduates of the Massachusetts Institute of Technology. The software enables developers to write and edit large sections of code and fix bugs using AI models from providers such as OpenAI, Anthropic, Google, and xAI. _🔗 Source: [The Economic Times](https://economictimes.indiatimes.com/tech/artificial-intelligence/meet-the-newest-billionaires-of-silicon-valley-founders-of-ai-coding-tool

Chinese President Xi Jinping told the World AI Conference in Shanghai that China will share AI technology with developing countries to avoid creating new global inequalities. He announced partnerships with Africa, Latin America, Asia, and BRICS nations to expand AI access. This signals China's push to position itself as a leader in global AI governance and development.

NVIDIA is launching two new, cheaper versions of its Jetson Thor computer boards for robots and edge AI devices (T3000 and T2000) arriving Q1 2027. The T3000 delivers strong AI performance at about half the power of the top model, while the T2000 handles lighter tasks. NVIDIA is also releasing memory-saving software tricks so customers don't need as much expensive memory. This matters because cheaper, more efficient edge AI hardware could make smart robots, drones, and industrial automation more affordable for businesses.

An AI startup called General Compute just secured a $400 million loan using inference chips as collateral — likely the first time chips specifically built to run AI models (not train them) have been used this way. The deal signals a shift in AI investment from flashy GPU financing toward cheaper, more efficient inference hardware. The startup uses SambaNova chips that claim 16x faster inference than GPU clouds while consuming less power. Investors see this as a move away from Nvidia's dominance as alternatives like Groq and Cerebras gain traction.

Chinese AI startup Moonshot released Kimi K3, the largest open-source AI model at 2.8 trillion parameters, which outperformed Anthropic's premium Fable 5 model on coding tasks while costing just $15 per million tokens versus Fable 5's $50. This reignites the debate over whether expensive American AI models are worth the price when open-source alternatives are catching up fast.

San Francisco's city attorney has ordered Apple and Google to remove 13 AI apps that create non-consensual nude images (deepfakes), saying the tech giants are profiting from sexual abuse. These 'nudify' apps use AI to digitally remove clothing from photos of people without their consent. Apple and Google say they've already deleted hundreds of such apps. The move signals growing legal pressure on app stores to police harmful AI tools.

Anthropic, the company behind the Claude AI chatbot, is reportedly planning an IPO as early as October, with Morgan Stanley, Goldman Sachs, and JPMorgan leading preparations. The timing is notable: China's Moonshot AI has just overtaken Claude on a coding benchmark with its Kimi K3 model, showing how quickly Chinese AI companies are closing the gap on US rivals. Venture capitalists say Chinese AI firms can run inference at a fraction of Western costs, adding pressure on Anthropic to defend its pricing once public.

Suno AI, a popular AI music generator, was hacked and the breach exposed that it scraped millions of songs from YouTube Music, Deezer, and Genius without permission. The hacker accessed customer data and payment records too. This confirms what music industry lawsuits alleged - Suno trained on 'tens of millions' of copyrighted songs, claiming it's 'fair use.' The leaked source code shows the company ingested over 113,000 hours from YouTube Music alone. If you've used AI music tools or care about artist rights, this matters because it shows how AI 'learns' from creators' work without asking.

Linus Torvalds, the creator of Linux, has publicly backed AI code review tools in the Linux kernel project, telling critics who oppose AI assistance to 'fork off' if they disagree. His comments came during a debate about Sashiko, an AI-powered code review tool from the Linux Foundation. Torvalds made clear he thinks AI tools are fine to use and he'll ignore anyone trying to persuade others not to use them.

Patreon is now actively blocking AI companies from scraping creator content for training their models. Working with Cloudflare, Patreon moved beyond simply asking AI bots to stay away (robots.txt requests) to directly blocking them. Testing showed thousands of weekly scraping attempts dropped to zero after implementing AI Crawl Control. Patreon says creators deserve control over how their work is used by AI companies, and the platform won't force creators to accept AI training just to grow their audience.

Apple briefly overtook Nvidia to become the world's most valuable company for the first time in 2026. While Nvidia rode the AI chip boom to dominance, investors are now favouring Apple because of its lighter spending approach and promising AI features in upcoming products. Nvidia slipped just 9% this year while Apple surged 23%.

Enterprise software giants like Workday, Freshworks, and Snowflake are bracing for disruption as AI agents increasingly handle tasks that users previously did through traditional apps. Gartner estimates $234 billion in SaaS spending is at risk from 'agentic arbitrage' by 2030. However, executives argue the threat is overblown—companies with trusted data platforms and deep domain expertise will survive by becoming the 'durable capabilities' that AI agents still need to access.

Apple[filed a trade secrets lawsuit against OpenAI](https://techcrunch.com/2026/07/13/the-wildest-allegations-in-apples-trade-secrets-lawsuit-against-openai/)last Friday, and it’s not messing around. The complaint alleges a pattern of misconduct reaching all the way up to OpenAI’s chief hardware officer and claims more than 400 former Apple employees now work at the company.[OpenAI’s response so far](https://techcrunch.com/2026/07/14/openai-pushes-back-on-apple-trade-secret-lawsuit/)has been carefully hedged, and the timing couldn’t be worse with the company reportedly eyeing an IPO as early as later this year. On this episode of TechCrunch’s[Equity](https://techcrunch.com/podcasts/equity/)podcast, hosts Kirsten Korosec, Anthony Ha, and Sean O’Kane dig into what the lawsuit could mean for

 Prakhar Khanna/ZDNET _Follow ZDNET:_[Add us as a preferred source](https://cc.zdnet.com/v1/otc/00hQi47eqnEWQ6T9d4QLBUc?element=BODY&element_label=Add+us+as+a+preferred+source&module=LINK&object_type=text-link&object_uuid=bf3378db-4147-4684-99fb-28842ca7decb&position=1&template=article&track_code=__COM_CLICK_ID__&url=https%3A%2F%2Fwww.google.com%2Fpreferences%2Fsource%3Fq%3Dzdnet.com&view_instance_uuid=6242d031-6519-4c84-a7d5-43bcfd535fa3&object_version=2e469f9e-90a7-41df-b71b-937c2cff09dd)_on Google._ * * * ### ZDNET's key takeaways * iOS 27 brings some much-needed changes to the Cont

NVIDIA and Hugging Face have released a guide on scaling AI model fine-tuning using NVIDIA NeMo Automodel combined with the Diffusers library. The tutorial covers how developers can efficiently train video and image generation models on large datasets.

Unlock world-class roadmaps & trusted Bench data.See More × ## Unparalleled insights. Industry analysis. Insider access. **Tom's Hardware** Premium equips you with world-class coverage and detailed insights into the evolving hardware landscape. * ✓**Full access to our trusted Bench database:** Access granular performance data instantly. * ✓**Exclusive hardware roadmaps:** Peer into the future of the hardware industry. * ✓**Daily news analysis:** Dive deep into the biggest stories. [Subscribe to our annual plan for just $29](https://www.tomshardware.com/subscription) Join Tom’s Hardware today Upgrade to Tom’s Hardware Premium [Explore](https://www.tomshardware.com/membership)[GO PREMIUM](https://www.tomshardware.com/tech-industry/policy/ai-data-centers-must-produce-as-much-power-as-they-use

_This is today's edition of_[_The Download_](https://forms.technologyreview.com/newsletters/briefing-the-download/?_ga=2.179569122.736533416.1649661040-405833893.1649413289),_our weekday newsletter that provides a daily dose of what's going on in the world of technology._ ### **There’s a lot of hype around perimenopause. Don’t buy it.** Perimenopause used to be considered taboo, but not anymore. Thanks at least in part to TV doctors and social media influencers, conversations about the sometimes years-long period before menopause are now more open than ever. But the conversation is increasingly shaped by misinformation. Despite what some marketers will claim, there is no test for perimenopause. That doesn’t mean women should have to put up with symptoms, but treatment suggestions often lac

A month-long test compared three smartwatches for blood pressure monitoring: Samsung Galaxy Watch Ultra, Amazfit Cheetah 2 Ultra, and YHE BP Doctor Fit. The Amazfit Cheetah 2 Ultra came closest to a clinical arm cuff, staying within 2-4 mmHg accuracy. Samsung's readings were 12-14 mmHg off, while YHE was most inconsistent.

 Adrian Kingsley-Huges/ZDNET _Follow ZDNET:_[Add us as a preferred source](https://cc.zdnet.com/v1/otc/00hQi47eqnEWQ6T9d4QLBUc?element=BODY&element_label=Add+us+as+a+preferred+source&module=LINK&object_type=text-link&object_uuid=4c096865-ba6b-4c25-be95-b26e124254a9&position=1&template=article&track_code=__COM_CLICK_ID__&url=https%3A%2F%2Fwww.google.com%2Fpreferences%2Fsource%3Fq%3Dzdnet.com&view_instance_uuid=48927f40-048a-4c1c-b2c4-aa4f808727df&object_version=8b1cea68-55ca-4896-bba4-67c49e08c019)_on G

 Kerry Wan/ZDNET _Follow ZDNET:[Add us as a preferred source](https://cc.zdnet.com/v1/otc/00hQi47eqnEWQ6T9d4QLBUc?element=BODY&element\_label=Add+us+as+a+preferred+source&module=LINK&object\_type=text-link&object\_uuid=d35d5655-d3cf-4483-b793-ade70b5c0ace&position=1&template=article&track\_code=\_\_COM\_CLICK\_ID\_\_&url=https%3A%2F%2Fwww.google.com%2Fpreferences%2Fsource%3Fq%3Dzdnet.com&view\_instance\_uuid=3335e109-815a-4495-8c2b-2248ead97771&object\_version=7f692ff6-f760-4fc5-a762-508e35990be4)on Google._ * * * ### ZDNET's key takeaways * Tax-free shopping typically runs in late July and early August. * Only

TechCrunch is promoting a panel at its Disrupt 2026 event focused on helping pre-seed founders secure funding when they have no product to show. The session features VCs discussing how AI is compressing development timelines while raising the bar for early-stage fundraising. The article profiles three speakers from Axiom Partners, True Ventures, and Slauson & Co. The content is primarily promotional/event coverage.

Tencent is testing Xiaowei, a new AI assistant built into WeChat, with a small group of users in China. The tool runs on Tencent's WeLM large language model and sometimes DeepSeek, letting users complete tasks like ordering food or booking rides directly through WeChat's mini-program ecosystem. With 1.38 billion monthly active users across WeChat and Weixin, the commercial reach could be massive if rolled out widely.

Google reportedly told Meta it could not meet the social media giant's full request for Gemini AI computing capacity around March. Meta's internal AI projects were delayed as a result, and staff were asked to use AI tokens more efficiently. Google CEO Sundar Pichai previously acknowledged near-term capacity constraints, with Google Cloud's backlog doubling to $462 billion. Meta is planning to spend up to $135 billion on AI infrastructure in 2026, deploying millions of chips across its data centers.

_Welcome to Tech in Asia’s daily newsletter, your essential dose of Asia’s tech and startup buzz. Not yet a subscriber? [Register here](https://www.techinasia.com/subscription?utm\_source=web&utm\_medium=article&utm\_campaign=dailynewsletter). Got a story tip? Send it to [email protected]._ ## In focus * Marketplace fees [force enablers to adapt](https://www.techinasia.com/rising-fees-changing-role-sea-ecommerce-enablers) * Workers [get paid to train robots](https://www.techinasia.com/robot-data-gold-rush-borrowed-time) * Tracking [SEA’s funding slowdown](https://www.techinasia.com/visual-story/capital-sea-southeast-asias-funding-landscape) * * * Hello reader, I usually buy PS5 games on ecommerce platforms, but over the past few months, several of the stores I regularly buy from have

[Cursor](https://cursor.com/), an AI-powered coding tool, has raised US$2.3 billion from investors including Nvidia, Google, Accel, Thrive Capital, and Andreessen Horowitz at a US$29.3 billion valuation. Forbes estimates that cofounders Michael Truell, Aman Sanger, Sualeh Asif, and Arvid Lunnemark now have a combined 4.5% stake in the company, worth at least US$1.3 billion. Cursor was founded in 2022 by four graduates of the Massachusetts Institute of Technology. The software enables developers to write and edit large sections of code and fix bugs using AI models from providers such as OpenAI, Anthropic, Google, and xAI. _🔗 Source: [The Economic Times](https://economictimes.indiatimes.com/tech/artificial-intelligence/meet-the-newest-billionaires-of-silicon-valley-founders-of-ai-coding-tool

OpenAI and Instacart have launched a feature letting users browse groceries, get AI-powered recommendations, and checkout directly within ChatGPT. Instacart handles delivery. This is the first app to use OpenAI's agentic commerce protocol for direct checkout, signalling a shift toward AI handling more everyday shopping tasks.

Micron Technology is pouring US$24 billion into Singapore over the next decade to build a new NAND memory chip factory. The facility, expected to start producing wafers in the second half of 2028, will create about 1,600 local jobs. This is part of Micron's global push to address a memory chip shortage driven by surging AI infrastructure demand.

Singapore's government is doubling down on construction robots and automation. National Development Minister Chee Hong Tat announced free sandbox spaces for testing construction tech, waived rental fees up to $100,000, and mandatory robots in JTC construction tenders since late 2025. Local AI firms like Vebits AI, Invigilo AI, and MyrLabs are already deploying computer vision and robotics for worksite monitoring and drain inspections. New regulations now let companies run smart hoists without operators inside, saving manpower.

Chinese AI startup Moonshot just launched Kimi K3, the world's largest open-source AI model with 2.8 trillion parameters. It can process one million tokens at once and performs competitively with top US models like Anthropic's Fable and OpenAI's GPT-5.5. The model is designed for complex coding and reasoning tasks. Moonshot, backed by Alibaba and Tencent, is reportedly raising $2 billion ahead of a potential Hong Kong listing.

Chinese President Xi Jinping told the World AI Conference in Shanghai that China will share AI technology with developing countries to avoid creating new global inequalities. He announced partnerships with Africa, Latin America, Asia, and BRICS nations to expand AI access. This signals China's push to position itself as a leader in global AI governance and development.

Malaysian company Zetrix AI is developing an AI avatar of Prime Minister Anwar Ibrahim and a digital identification platform for the Philippine government. The firm is positioning itself as a regional player combining AI avatars with Web3 technology.

Xi Jinping used his debut at China's World AI Conference to push for global AI cooperation, positioning China as a champion of affordable, open AI for developing nations. He's announced training programmes for 5,000 AI professionals from developing countries and partnerships across ASEAN, Africa, Latin America, and BRICS. Beijing is also spending 2 trillion yuan over five years to build its own AI infrastructure and reduce reliance on US chips. This matters because China is directly challenging American dominance in AI and offering countries a viable alternative to Western technology.

Chinese AI startup Moonshot AI is in early talks for a Hong Kong IPO and raising up to US$1 billion in new funding, potentially valuing the company at around US$18 billion. That's a fourfold jump from its US$4.8 billion valuation just months ago. The timing comes as other Hong Kong-listed AI firms like Z.ai and MiniMax have seen their market caps soar to four to five times their IPO prices, making this a hot moment for AI listings. Rival firms are burning hundreds of millions annually just to stay competitive in China's AI race.

A fake think-tank called the Institute of East Asia Strategic Studies (IEASS) claiming to be based in Singapore has been exposed for using AI-generated profiles to approach security analysts and intelligence professionals in Singapore, Taiwan, and other countries. The institute had no verifiable employees—some profile pictures showed telltale visual artefacts of AI-generated images, and supposed graduates from Singapore University of Social Sciences had no enrollment records. The think-tank was linked to two similar operations using the same website infrastructure built with AI platform Lovable. Experts say this is a classic intelligence-gathering tactic, with one noting major powers are "all doing this."

A podcast discussion featuring four-time founder Prashanth Ranganathan explores how AI has transformed startup building — from faster prototyping to new fundraising dynamics — while potentially creating pitfalls like mistaking speed for real customer understanding. The key insight: as AI makes execution cheaper, human judgment becomes more valuable.

TSMC reported May 2026 revenue of NT$416.98 billion (US$13.2 billion), up 30.1% year-on-year. The chipmaker attributed the growth to strong AI-related demand, with high-performance computing accounting for 61% of its Q1 revenue.

OpenAI is rolling out new safeguards for teen ChatGPT users, including an option for parents to enable a Study Mode for linked teen accounts. The changes aim to make the AI assistant safer for younger users.

Chinese President Xi Jinping told an AI conference in Shanghai that no single nation should dominate artificial intelligence, calling for global cooperation instead of US-led restrictions on technology access. He announced China would train 5,000 AI professionals from developing countries and expand AI partnerships across ASEAN, Africa, Latin America, and BRICS nations. Xi also criticised the 'overstretching' of national security concerns in AI - a direct jab at US chip export controls blocking China's access to advanced technology. The speech positions China as offering cheaper, open-source AI alternatives globally.

Sometimes, you have to dig beyond the headlines to get a better idea of how a company is performing. Endowus’ latest audited financial results are a good example of this. [Gregory Van](https://www.linkedin.com/in/gregory-van/?originalSubdomain=hk), the co-founder and CEO of the Singapore-based wealthtech firm, told _Tech in Asia_ last year that it had a [“clear path to profitability”](https://www.techinasia.com/endowus-doubles-revenue-2022-losses-rose-26m) by 2025. On its face, the results suggest that the firm has its work cut out for it to hit this milestone. Endowus co-founders Samuel Rhee (left) and Gregory Van / Photo credit: Endowus In 2023, Endowus’ total losses before tax widened to S$30.6 million (US$23.7 million) – a 14% jump from last year. This came even as its revenue grew by

[Kioxia Holdings](https://www.kioxia-holdings.com/en-jp/top.html) shares fell over 12% after news that BCPE Pangea Cayman LP, a Bain Capital-backed entity, plans to sell 36 million shares to overseas investors. The stake sale may be valued at about ¥355 billion (US$2.3 billion), based on the November 25 closing price. Kioxia is a Japanese multinational computer memory manufacturer. Kioxia’s stock has surged in recent months on expectations for strong AI-related demand but remains well above its IPO price of ¥1,455 (US$9.31). The shares have dropped since the company reported quarterly earnings that missed market expectations earlier in November. _🔗 Source: [Bloomberg](https://www.bloomberg.com/news/articles/2025-11-26/chipmaker-kioxia-sinks-after-bain-backed-firm-discloses-selloff)_ ## 🧠 F

[News](https://www.techinasia.com/)Premium * [News](https://www.techinasia.com/) * [Jobs](https://www.techinasia.com/jobs/search) * [Companies](https://www.techinasia.com/companies) * [Events](https://www.techinasia.com/events/live) * [About](https://www.techinasia.com/about) * [advertise](https://www.techinasia.com/advertise?utm_source=techinasia.com&utm_medium=website&utm_campaign=intl-navbar) * [Tech in Asia Conference](https://conference.techinasia.com/?utm_source=navbar-tia-homepage&utm_medium=website&utm_campaign=conference-2026) * [Sign Up](https://www.techinasia.com/auth) * [Log In](https://www.techinasia.com/auth?ref=nav_login) [Subscribe](https://www.techinasia.com/subscription) [I'm fundraising](https://www.techinasia.com/companies/fundraising) [Subscribe](https://www.techinasia

A model released Friday by Chinese startup Moonshot AI has fuelled buzz around the country's tech prowess, as experts said it could rival some of the more advanced offerings from US labs. Large language models underpin chatbots and other artificial intelligence tools with their ability to crunch huge amounts of digital data. Moonshot AI's "Kimi K3" is one of several from China growing in global popularity thanks to their lower costs and source code that programmers can customise. Soon after its launch, Kimi K3 had topped a leaderboard for AI coding run by a platform called Arena created by UC Berkeley researchers. That drew excitement from industry insiders, with some evoking a 2025 release from China's DeepSeek that shook assumptions of US dominance in AI. "Kimi K3 seems really good, clos

A major cybersecurity report reveals that compromised logins — stolen passwords, hacked accounts — now drive 79% of ransomware attacks worldwide, surpassing software bugs as the top entry point for hackers. In Singapore, two-thirds (67%) of ransomware attacks hit organisations through stolen identities. The scary part: 97% of these identity breaches happened even when the victims had multi-factor authentication (MFA) enabled — meaning the extra security layer isn't foolproof. Recovery costs average $1.7 million per incident globally. Experts warn that as AI tools get smarter, hackers can now scan for weak spots in login systems faster and cheaper than before. Bottom line: your organisation's login security needs an urgent upgrade.

Standard Chartered has moved 70% of its global infrastructure onto a software-defined private cloud built on VMware Cloud Foundation, cutting deployment time from weeks down to a single day. The bank, operating across 54 markets, is using this modernization to embed zero-trust security directly into its infrastructure and support always-on banking services. The partnership positions the bank for further AI-driven innovation in financial services.

Indonesia is overhauling its copyright laws in ways that could significantly impact how AI companies like Google operate. The rewrite targets how tech platforms use copyrighted content for training AI models and displaying search results. If passed, companies may need to negotiate licensing deals with content creators or face legal action — a move that could reshape AI development in Southeast Asia and increase costs for major platforms.

The Public Service Division clarifies that GovTech's job cuts are not part of a wider government-wide restructuring. Basically, the 93 layoffs (and future cuts) are specific to GovTech's shift to a new operating model, not a signal that other government agencies will follow.

Chinese AI startup Moonshot released Kimi K3, an open AI model that ranks just behind OpenAI's GPT-5.6 and Anthropic's Claude Fable 5 — beating everything else including other US models. The company priced it at premium rates, signaling China's AI race has moved beyond cheap competition to quality-focused differentiation. The launch spooked rival Chinese AI stocks, which fell sharply on competition concerns.

Nvidia is building a 140-megawatt AI data center in Japan and partnering with Japanese manufacturers on AI-powered robotics for factories. The project covers everything from national AI infrastructure to quantum computing research.

Chinese chip and cloud companies are on a tear in 2025, with stocks like Cambricon Technologies surging 87% and Semiconductor Manufacturing International up 83% year-to-date. The rally is fueled by Beijing's push for tech self-reliance and government pressure on firms to use domestic chips over Nvidia's offerings. The SSE Science and Technology Innovation Board 50 Index gained 28% in August alone — its biggest monthly gain ever. However, some fund managers are warning that valuations are stretched, with chip stocks trading at significant premiums to historical averages.

Chinese AI companies are forming alliances to build their own tech ecosystem without relying on American chips and software. At Shanghai's World AI Conference, StepFun launched partnerships connecting AI model developers with chip makers like Huawei and Biren — both hit by US sanctions. Another group brings together SenseTime, MiniMax, and Metax to push AI into factories and businesses.

A working mom describes how she uses AI agents to juggle childcare, family calendars, and daily mental load. The piece offers a practical look at how AI tools can help manage real-life parenting chaos — scheduling pickups, remembering appointments, and handling the endless mental checklist of running a household.

Researchers have developed robotic clothing designed to help people dress themselves — particularly useful for elderly or disabled individuals who struggle with buttons, zippers, and tight sleeves. The system uses sensors and motors to adjust fabric tension and guide limbs through clothing. It represents a practical intersection of robotics and daily life assistance, though current prototypes are bulky and expensive.

San Francisco's city attorney has ordered Apple and Google to remove 13 AI apps that create non-consensual nude images (deepfakes), saying the tech giants are profiting from sexual abuse. These 'nudify' apps use AI to digitally remove clothing from photos of people without their consent. Apple and Google say they've already deleted hundreds of such apps. The move signals growing legal pressure on app stores to police harmful AI tools.

Cybersecurity firm Tenable has updated its platform to help companies track security vulnerabilities in code as AI tools let developers ship software much faster. A recent study found generative AI can speed up code output by 3-4x but introduces flaws at 10x the normal rate. The new integration connects code vulnerability data with broader security monitoring, letting teams prioritise real risks instead of drowning in alerts.

NVIDIA is launching two new, cheaper versions of its Jetson Thor computer boards for robots and edge AI devices (T3000 and T2000) arriving Q1 2027. The T3000 delivers strong AI performance at about half the power of the top model, while the T2000 handles lighter tasks. NVIDIA is also releasing memory-saving software tricks so customers don't need as much expensive memory. This matters because cheaper, more efficient edge AI hardware could make smart robots, drones, and industrial automation more affordable for businesses.

Adding AI to a business doesn't automatically make it more efficient — companies need to redesign their entire operating model, not just bolt on new technology. That's the argument from enterprise tech analyst Robert Kramer, who says most organizations still have slow decisions, disputed data, and manual handoffs despite using AI. Top performers invest heavily in data quality and redesign workflows from start to finish, with senior leaders owning AI governance.

Linus Torvalds, the creator of Linux, has publicly backed AI code review tools in the Linux kernel project, telling critics who oppose AI assistance to 'fork off' if they disagree. His comments came during a debate about Sashiko, an AI-powered code review tool from the Linux Foundation. Torvalds made clear he thinks AI tools are fine to use and he'll ignore anyone trying to persuade others not to use them.

An AI startup called General Compute just secured a $400 million loan using inference chips as collateral — likely the first time chips specifically built to run AI models (not train them) have been used this way. The deal signals a shift in AI investment from flashy GPU financing toward cheaper, more efficient inference hardware. The startup uses SambaNova chips that claim 16x faster inference than GPU clouds while consuming less power. Investors see this as a move away from Nvidia's dominance as alternatives like Groq and Cerebras gain traction.

GovTech held an 'Ask Me Anything' session with staff on July 16 after announcing plans to cut 7-9% of its workforce — the first large-scale government retrenchment in years. Around 200 staff attended in person. Morale remains low with widespread anxiety about job security. The first phase affects 93 staff; about 110 more will be retrained and redeployed. Those hit hardest are in their 40s and 50s.

OpenAI's Codex coding tool has been deleting users' files without permission after a recent update to GPT-5.6 Sol. Engineers reported losing entire production databases and Mac folders. OpenAI confirmed the behavior happens when users give the system full computer access without safeguards enabled. The company is now adding warnings and safety checks, though it says the incidents are 'extremely rare.'

Netflix has admitted to using generative AI in roughly 300 films and TV shows, primarily in post-production for complex scenes like crowds, battle sequences, and establishing shots. The company says AI allows them to include key scenes that would otherwise be cut and deliver higher quality faster and at lower cost. The disclosure came amid Netflix's shareholder results showing slowing revenue growth, as the company faces pressure to reinvent itself after losing co-CEO Reed Hastings.

Anthropic, the company behind the Claude AI chatbot, is reportedly planning an IPO as early as October, with Morgan Stanley, Goldman Sachs, and JPMorgan leading preparations. The timing is notable: China's Moonshot AI has just overtaken Claude on a coding benchmark with its Kimi K3 model, showing how quickly Chinese AI companies are closing the gap on US rivals. Venture capitalists say Chinese AI firms can run inference at a fraction of Western costs, adding pressure on Anthropic to defend its pricing once public.

_This is today's edition of_[_The Download_](https://forms.technologyreview.com/newsletters/briefing-the-download/?_ga=2.179569122.736533416.1649661040-405833893.1649413289),_our weekday newsletter that provides a daily dose of what's going on in the world of technology._ ### **There’s a lot of hype around perimenopause. Don’t buy it.** Perimenopause used to be considered taboo, but not anymore. Thanks at least in part to TV doctors and social media influencers, conversations about the sometimes years-long period before menopause are now more open than ever. But the conversation is increasingly shaped by misinformation. Despite what some marketers will claim, there is no test for perimenopause. That doesn’t mean women should have to put up with symptoms, but treatment suggestions often lac

GovTech is cutting 7-9% of its workforce over two years, eliminating vendor management and project delivery roles while hiring more in-house engineers and cross-functional builders. The agency says AI was not the driver, but acknowledges that rapid software development cycles now demand leaner, more technical teams. This reflects a broader shift in tech workplaces toward agile, outcome-driven roles with direct ownership of products.

AI companies are borrowing massive amounts to build data centers — Morgan Stanley projects $570 billion in AI-related debt in 2026, four times last year's pace. But don't panic: Big Tech isn't broke. They're choosing to spread costs across credit markets rather than burning cash. The real shift: risk is moving from public bonds into private credit and off-balance-sheet deals like Meta's $27 billion Hyperion joint venture, where outside investors now own 80% of the project.

Suno AI, a popular AI music generator, was hacked and the breach exposed that it scraped millions of songs from YouTube Music, Deezer, and Genius without permission. The hacker accessed customer data and payment records too. This confirms what music industry lawsuits alleged - Suno trained on 'tens of millions' of copyrighted songs, claiming it's 'fair use.' The leaked source code shows the company ingested over 113,000 hours from YouTube Music alone. If you've used AI music tools or care about artist rights, this matters because it shows how AI 'learns' from creators' work without asking.

Apple briefly overtook Nvidia to become the world's most valuable company for the first time in 2026. While Nvidia rode the AI chip boom to dominance, investors are now favouring Apple because of its lighter spending approach and promising AI features in upcoming products. Nvidia slipped just 9% this year while Apple surged 23%.

Enterprise software giants like Workday, Freshworks, and Snowflake are bracing for disruption as AI agents increasingly handle tasks that users previously did through traditional apps. Gartner estimates $234 billion in SaaS spending is at risk from 'agentic arbitrage' by 2030. However, executives argue the threat is overblown—companies with trusted data platforms and deep domain expertise will survive by becoming the 'durable capabilities' that AI agents still need to access.

Apple[filed a trade secrets lawsuit against OpenAI](https://techcrunch.com/2026/07/13/the-wildest-allegations-in-apples-trade-secrets-lawsuit-against-openai/)last Friday, and it’s not messing around. The complaint alleges a pattern of misconduct reaching all the way up to OpenAI’s chief hardware officer and claims more than 400 former Apple employees now work at the company.[OpenAI’s response so far](https://techcrunch.com/2026/07/14/openai-pushes-back-on-apple-trade-secret-lawsuit/)has been carefully hedged, and the timing couldn’t be worse with the company reportedly eyeing an IPO as early as later this year. On this episode of TechCrunch’s[Equity](https://techcrunch.com/podcasts/equity/)podcast, hosts Kirsten Korosec, Anthony Ha, and Sean O’Kane dig into what the lawsuit could mean for

 Prakhar Khanna/ZDNET _Follow ZDNET:_[Add us as a preferred source](https://cc.zdnet.com/v1/otc/00hQi47eqnEWQ6T9d4QLBUc?element=BODY&element_label=Add+us+as+a+preferred+source&module=LINK&object_type=text-link&object_uuid=bf3378db-4147-4684-99fb-28842ca7decb&position=1&template=article&track_code=__COM_CLICK_ID__&url=https%3A%2F%2Fwww.google.com%2Fpreferences%2Fsource%3Fq%3Dzdnet.com&view_instance_uuid=6242d031-6519-4c84-a7d5-43bcfd535fa3&object_version=2e469f9e-90a7-41df-b71b-937c2cff09dd)_on Google._ * * * ### ZDNET's key takeaways * iOS 27 brings some much-needed changes to the Cont

Unlock world-class roadmaps & trusted Bench data.See More × ## Unparalleled insights. Industry analysis. Insider access. **Tom's Hardware** Premium equips you with world-class coverage and detailed insights into the evolving hardware landscape. * ✓**Full access to our trusted Bench database:** Access granular performance data instantly. * ✓**Exclusive hardware roadmaps:** Peer into the future of the hardware industry. * ✓**Daily news analysis:** Dive deep into the biggest stories. [Subscribe to our annual plan for just $29](https://www.tomshardware.com/subscription) Join Tom’s Hardware today Upgrade to Tom’s Hardware Premium [Explore](https://www.tomshardware.com/membership)[GO PREMIUM](https://www.tomshardware.com/tech-industry/policy/ai-data-centers-must-produce-as-much-power-as-they-use

 Kerry Wan/ZDNET _Follow ZDNET:[Add us as a preferred source](https://cc.zdnet.com/v1/otc/00hQi47eqnEWQ6T9d4QLBUc?element=BODY&element\_label=Add+us+as+a+preferred+source&module=LINK&object\_type=text-link&object\_uuid=d35d5655-d3cf-4483-b793-ade70b5c0ace&position=1&template=article&track\_code=\_\_COM\_CLICK\_ID\_\_&url=https%3A%2F%2Fwww.google.com%2Fpreferences%2Fsource%3Fq%3Dzdnet.com&view\_instance\_uuid=3335e109-815a-4495-8c2b-2248ead97771&object\_version=7f692ff6-f760-4fc5-a762-508e35990be4)on Google._ * * * ### ZDNET's key takeaways * Tax-free shopping typically runs in late July and early August. * Only

A month-long test compared three smartwatches for blood pressure monitoring: Samsung Galaxy Watch Ultra, Amazfit Cheetah 2 Ultra, and YHE BP Doctor Fit. The Amazfit Cheetah 2 Ultra came closest to a clinical arm cuff, staying within 2-4 mmHg accuracy. Samsung's readings were 12-14 mmHg off, while YHE was most inconsistent.

 Adrian Kingsley-Huges/ZDNET _Follow ZDNET:_[Add us as a preferred source](https://cc.zdnet.com/v1/otc/00hQi47eqnEWQ6T9d4QLBUc?element=BODY&element_label=Add+us+as+a+preferred+source&module=LINK&object_type=text-link&object_uuid=4c096865-ba6b-4c25-be95-b26e124254a9&position=1&template=article&track_code=__COM_CLICK_ID__&url=https%3A%2F%2Fwww.google.com%2Fpreferences%2Fsource%3Fq%3Dzdnet.com&view_instance_uuid=48927f40-048a-4c1c-b2c4-aa4f808727df&object_version=8b1cea68-55ca-4896-bba4-67c49e08c019)_on G

TechCrunch is promoting a panel at its Disrupt 2026 event focused on helping pre-seed founders secure funding when they have no product to show. The session features VCs discussing how AI is compressing development timelines while raising the bar for early-stage fundraising. The article profiles three speakers from Axiom Partners, True Ventures, and Slauson & Co. The content is primarily promotional/event coverage.

Moonshot AI's Kimi K3 model has jumped from 17th place to number one on AI coding benchmarks, beating top US models like Claude and GPT in frontend web development tasks. The catch: Kimi is open-weight, meaning anyone can inspect its code and use it cheaply — roughly six times less than comparable closed models. Investors are nervous as the news coincides with a US tech stock selloff, raising fresh questions about whether American AI companies can justify their premium pricing when Chinese rivals are matching performance at a fraction of the cost.

Chinese AI startup Moonshot released Kimi K3, the largest open-source AI model at 2.8 trillion parameters, which outperformed Anthropic's premium Fable 5 model on coding tasks while costing just $15 per million tokens versus Fable 5's $50. This reignites the debate over whether expensive American AI models are worth the price when open-source alternatives are catching up fast.

Patreon is now actively blocking AI companies from scraping creator content for training their models. Working with Cloudflare, Patreon moved beyond simply asking AI bots to stay away (robots.txt requests) to directly blocking them. Testing showed thousands of weekly scraping attempts dropped to zero after implementing AI Crawl Control. Patreon says creators deserve control over how their work is used by AI companies, and the platform won't force creators to accept AI training just to grow their audience.

NVIDIA and Hugging Face have released a guide on scaling AI model fine-tuning using NVIDIA NeMo Automodel combined with the Diffusers library. The tutorial covers how developers can efficiently train video and image generation models on large datasets.

Chinese AI startup Moonshot has launched Kimi K3, a 2.8-trillion-parameter model it claims can match top US AI systems from OpenAI and Anthropic. The model will be released as open-source on July 27, making it the world's first freely available AI in the three-trillion-parameter class. Independent benchmarks show it ranking highly in coding and web interface tasks. The release triggered sharp stock drops for Chinese competitors Zhipu and MiniMax, and revives fears similar to the DeepSeek moment — that Chinese labs can build AI matching Western capabilities without relying on restricted US hardware.

The UK's financial regulator (FCA) released a major report warning that AI is reshaping banking and finance—and making scams far more dangerous. AI-powered deepfakes, fake identities, and hyper-personalized phishing attacks are getting harder to spot, and criminals are moving faster than regulators. The report calls for stronger oversight as AI agents start handling money decisions autonomously.

South Korean AI chip startup FuriosaAI is ramping up production and developing a 2nm Stork chip optimized for AI inference tasks. This represents another challenger to NVIDIA and AMD in the AI chip space, which could eventually benefit companies looking to reduce their AI infrastructure costs.

Chinese President Xi Jinping told the World AI Conference in Shanghai that AI development should be a global effort, not a 'solo performance by any single country.' He announced a new World Artificial Intelligence Cooperation Organization with 29 countries, positioning it as China's answer to the US-led Pax Silica framework. Xi also pledged AI training for 5,000 people in developing countries and opposed export controls framed as national security. The move highlights growing geopolitical rivalry over AI governance.

Japanese companies are quietly reducing their reliance on Nvidia GPUs and exploring alternatives, with South Korean neural processing units (NPUs) emerging as a viable option. The shift comes as GPU costs stay high and companies look for more specialised, cost-effective chips for AI workloads.

X Trending * [I wore the world's first HDR10 smart glasses](https://www.zdnet.com/article/rayneo-air-4-pro-hdr10-smart-glasses-ces/) * [TCL's new E Ink tablet beats the Remarkable and Kindle](https://www.zdnet.com/article/tcl-note-a1-nxtpaper-ces/) * [Anker's new charger is one of the most unique I've ever seen](https://www.zdnet.com/article/anker-wireless-chargers-ces-2026/) * [Best laptop cooling pads](https://www.zdnet.com/home-and-office/best-laptop-cooling-pad/) * [Best flip phones](https://www.zdnet.com/article/best-flip-phone/) * [Roku TV vs Fire Stick](https://www.zdnet.com/home-and-office/home-entertainment/roku-tv-vs-fire-stick/) * [Galaxy Buds 3 Pro vs Apple AirPods Pro 3](https://www.zdnet.com/article/apple-airpods-pro-3-vs-samsung-galaxy-buds-3-pro/) * [M5 MacBook Pro vs M4 Ma

X Trending * [I wore the world's first HDR10 smart glasses](https://www.zdnet.com/article/rayneo-air-4-pro-hdr10-smart-glasses-ces/) * [TCL's new E Ink tablet beats the Remarkable and Kindle](https://www.zdnet.com/article/tcl-note-a1-nxtpaper-ces/) * [Anker's new charger is one of the most unique I've ever seen](https://www.zdnet.com/article/anker-wireless-chargers-ces-2026/) * [Best laptop cooling pads](https://www.zdnet.com/home-and-office/best-laptop-cooling-pad/) * [Best flip phones](https://www.zdnet.com/article/best-flip-phone/) * [Roku TV vs Fire Stick](https://www.zdnet.com/home-and-office/home-entertainment/roku-tv-vs-fire-stick/) * [Galaxy Buds 3 Pro vs Apple AirPods Pro 3](https://www.zdnet.com/article/apple-airpods-pro-3-vs-samsung-galaxy-buds-3-pro/) * [M5 MacBook Pro vs M4 Ma

A U.S. judge refused to block Meta from laying off 26 workers who are suing the company, claiming Meta used AI tools to select jobs for cuts in ways that discriminated against employees with disabilities or those on medical leave. The lawsuit — believed to be the first major U.S. case challenging AI-powered layoffs — alleges Meta's systems penalized workers whose AI adoption scores dropped while on protected leave. The workers lose health insurance and stock options. Meta denies wrongdoing, saying humans made the final decisions.

BMW has become the first carmaker to let shoppers configure and buy vehicles directly through ChatGPT. Instead of clicking through menus on BMW's website, users can describe the car they want — trim, colour, options — in plain text or voice. The move puts BMW's sales funnel inside a third-party AI assistant, potentially reaching buyers who use ChatGPT for everyday tasks. The auto industry is watching to see if conversational shopping can shift big-ticket purchasing behaviour.

Apple has filed a trade secrets lawsuit against OpenAI, alleging the company poached over 400 former Apple employees and engaged in misconduct reaching up to OpenAI's chief hardware officer. The lawsuit could seriously complicate OpenAI's reported plans for an IPO later this year, as it raises questions about the company's practices and trustworthiness. This is a major escalation in the increasingly tense relationship between the two tech giants.

 Credit: CFOTO/Contributor, Future Publishing via Getty Images _Follow ZDNET:_[Add us as a preferred source](https://cc.zdnet.com/v1/otc/00hQi47eqnEWQ6T9d4QLBUc?element=BODY&element_label=Add+us+as+a+preferred+source&module=LINK&object_type=text-link&object_uuid=4994a1b8-e80f-433c-ba9d-e05e4758db72&position=1&template=article&track_code=__COM_CLICK_ID__&url=https%3A%2F%2Fwww.google.com%2Fpreferences%2Fsource%3Fq%3Dzdnet.com&view_instance_uuid=8b427668-207f-48cb-9efb-44c20e2d60a9&object_version=16f8793e-3257-4d36-866a-5a86179f0614)_on Google._ * * * AI labs are shipping new models nonsto

Apple has filed a trade secrets lawsuit against OpenAI, alleging the company recruited over 400 former Apple employees and engaged in systematic misconduct reaching up to OpenAI's chief hardware officer. The timing is painful for OpenAI — the company is reportedly eyeing an IPO later this year, and this legal battle could throw cold water on those plans. For everyday readers, this matters because it signals that big tech companies are getting serious about protecting their talent pipelines and trade secrets in the AI race.

iOS 27 is shaping up to be a polish update rather than a flashy overhaul. Five underrated features stand out: a redesigned Control Center with one-tap access to connectivity toggles, a new 'Captured by Me' folder in Photos for your own shots, smarter auto-paste suggestions above the keyboard, improved voice dictation that removes 'umms' and adds punctuation automatically using Apple's on-device AI model, and separate volume controls for ringer, alarms, and system sounds. These small quality-of-life tweaks add up in daily use, and they're available now in the public beta.

OpenAI's GPT-5.6 accidentally deleted user files when given full computer access. The AI overwrote a system variable and performed destructive actions without asking for confirmation. OpenAI is rolling out extra safeguards and investigating what went wrong.

A new security framework highlights how AI data centers face unique and severe risks that are outpacing the industry's ability to address them. GPU clusters storing models worth hundreds of millions, training pipelines, and inference systems are prime targets for attackers seeking intellectual property theft, data poisoning, or strategic disruption. Real incidents include a US tech company's AI datacenter being breached with IP stolen, a former Google engineer convicted of stealing AI datacenter designs for China, and over 10 petabytes of data reportedly stolen from China's Tianjin supercomputing hub. The structural problem: GPU scarcity means customers accept security risks they'd never tolerate elsewhere.

Palantir CEO Alex Karp says AI will make him roughly 20 times wealthier (potentially $300 billion), but predicts middle-class workers will only see modest salary increases over the next decade. He called the wealth gap a 'complete decoupling of unimaginable wealth and normal wealth' and criticized tech leaders for overselling AI's benefits to everyone. The remarks add to growing warnings from figures like Larry Fink and Geoffrey Hinton that AI gains are concentrating among a tiny group of owners while leaving most workers behind.

Amazon's AWS cloud service accidentally showed some customers bills for billions — or even trillions — of dollars, when their actual monthly spend was just a few cents. One user saw $7.1 trillion (yes, trillion) on their dashboard. AWS says it's a billing system error and no one actually owes the money, but the glitch caused genuine panic among customers who thought they'd suddenly be on the hook for billions.

TikTok is testing a tool that spots AI-generated copies of creators' faces and voices, letting them report the fakes directly to the platform. Deepfakes of real people are becoming harder to spot, so this gives creators some control over how their likeness is being used without permission.

SoftBank CEO Masayoshi Son dismissed critics of AI valuations as 'too stupid' to understand the technology's potential, doubling down on his firm's massive AI investments despite market skepticism.

Healthcare company Clover Health has disclosed that employee accounts were accessed in a cyber incident. The company says it's investigating but hasn't revealed how many accounts were affected or what data was accessed. Healthcare companies are frequent targets for cyberattacks because of the sensitive patient data they hold.

Your next smartphone might cost more — and it's because of AI. Tech companies building AI data centers are hogging memory chips (RAM and storage), leaving less capacity for phones and driving up prices. India just saw its biggest smartphone sales drop in six years, with budget phone shipments plummeting 45%. Prices in India have jumped 4% to 68% depending on the model. If you're in the market for an affordable phone, expect to pay more or wait longer before upgrading.

Australia's Prime Minister announced new national AI framework requirements: data centers must produce as much power as they consume, plus meet water efficiency standards and protect intellectual property rights. This comes as AI data centers draw growing opposition from communities worried about energy use, water consumption, and environmental damage — with 70% of Americans now opposing AI data centers in their areas.

Agility Robotics, the company behind the Digit humanoid robot, just opened a massive 60,000-square-foot training facility in Fremont, California—right up the road from Tesla's factory where its Optimus robots are expected to roll out this year. Unlike Tesla's Optimus, which is still in development, Agility's robots are already working: Digit has moved over 100,000 totes at logistics facilities for Amazon and GXO, and the company has locked in $300 million in contract orders. It's now training the next generation of its robot for tasks like unloading trucks and handling cardboard. The move puts Agility head-to-head with Tesla in the humanoid robot race, though both companies face the same challenge: convincing the world that bipedal robots can work safely alongside humans.

Disney is reportedly using AI-generated content in children's programming, raising concerns about creative quality and the displacement of human artists in family entertainment.

Chinese President Xi Jinping called for more open-source AI development at the World Artificial Intelligence Conference in Shanghai, saying China is ready to be "more open." His speech outlined four principles, including a push for global AI regulation while criticizing US "protectionism" through export controls. Many Chinese AI models—including DeepSeek, Kimi K3, and GLM—are already open-source, while American labs like OpenAI and Anthropic keep their models closed. Xi also pledged 5,000 AI training opportunities for developing nations.

Meta is reportedly in talks to rent out excess computing power from its data centers to Anthropic, one of its biggest AI rivals. The deal would let Anthropic access additional AI training and inference capacity while Meta monetizes infrastructure it built for its own AI ambitions. It's a sign that the massive spending on AI data centers might eventually pay off in unexpected ways—as a service other companies want to buy.

Singapore is actively developing strategies to help its workforce adapt as AI reshapes industries. The government is focusing on upskilling workers, updating education systems, and ensuring workers can transition into new roles as automation takes over routine tasks. This human capital strategy aims to keep Singapore competitive in the AI era while protecting workers from being left behind.

Vint Cerf, one of the internet's original architects, is joining an effort called DNSid to give AI agents permanent identity badges tied to existing internet infrastructure. The goal: when an AI agent misbehaves or causes damage, organizations can trace it back to whoever created it. Think of it like a license plate for AI—not a background check, just a way to establish accountability. The project aims to solve a growing problem as AI agents increasingly act autonomously across different companies and systems.

Tech executives are hiring armed bodyguards amid growing public backlash against AI. The anger stems from concerns about job losses, privacy issues, and existential risks from artificial intelligence. This reflects deepening divide between the AI industry and everyday people feeling its effects.

Data analytics company Databricks has raised approximately $3 billion in a new funding round, pushing its valuation to $188 billion—nearly tripling its value in under a year. The company has successfully transformed from a 'big data' era software firm into an AI powerhouse, now offering AI database products and championing affordable open-weight AI models for enterprise coding tasks. Its rapid valuation growth reflects how companies are desperately buying into the AI hype.

Luxury phone maker Vertu is selling a $6,880 foldable aimed at executives, with an AI agent called Hermes designed to handle real work tasks like trip planning, document analysis, and scheduling. Testing revealed the AI is ambitious but inconsistent—it can automate multi-step workflows but sometimes delivers incorrect results, like scheduling flights on wrong dates or setting reminders at bizarre times. The actual hardware underneath is essentially a rebranded ZTE phone with luxury materials slapped on.